Stimulation-induced increases in cerebral blood flowand local capillary vasoconstriction depend onconducted vascular responsesChangsi Caia, Jonas C. Fordsmanna, Sofie H. Jensenb, Bodil Gessleina, Micael Lønstrupa, Bjørn O. Halda,Stefan A. Zambacha, Birger Brodinb, and Martin J. Lauritzena,1

aCenter for Neuroscience, Faculty of Health and Medical Science, University of Copenhagen, 2200 Copenhagen N, Denmark; and bDepartment of Pharmacy,University of Copenhagen, 2200 Copenhagen N, Denmark

Edited by Marcus E. Raichle, Washington University in St. Louis, St. Louis, MO, and approved May 10, 2018 (received for review May 10, 2017)

Functional neuroimaging, such as fMRI, is based on coupling neu-ronal activity and accompanying changes in cerebral blood flow(CBF) and metabolism. However, the relationship between CBF andevents at the level of the penetrating arterioles and capillaries is notwell established. Recent findings suggest an active role of capillariesin CBF control, and pericytes on capillaries may be major regulatorsof CBF and initiators of functional imaging signals. Here, using two-photon microscopy of brains in living mice, we demonstrate thatstimulation-evoked increases in synaptic activity in the mousesomatosensory cortex evokes capillary dilation starting mostly atthe first- or second-order capillary, propagating upstream anddownstream at 5–20 μm/s. Therefore, our data support an activerole of pericytes in cerebrovascular control. The gliotransmitter ATPapplied to first- and second-order capillaries by micropipette puff-ing induced dilation, followed by constriction, which also propa-gated at 5–20 μm/s. ATP-induced capillary constriction was blockedby purinergic P2 receptors. Thus, conducted vascular responses incapillaries may be a previously unidentified modulator of cerebro-vascular function and functional neuroimaging signals.

conducted vascular responses | pericytes | neurovascular coupling |purinergic signaling | cerebral capillaries

Brain function emerges from signaling in and between neuronsand astrocytes, causing fluctuations in the cerebral metabolic

rate of oxygen and cerebral blood flow (CBF). Normal brainfunction depends on a preserved supply of glucose and oxygen,which is mediated by neurovascular coupling, the robust couplingbetween brain activity and CBF. Neurovascular coupling de-pends on the functional properties of the association of brainmicrovessels, astrocytes, pericytes, and neurons, which togetherconstitute the neurovascular unit (1).Brain arterioles are traditionally thought to control CBF and

brain capillaries to serve in the exchange of substances betweenthe blood and brain. This view of CBF dynamics was revolu-tionized recently by the discovery that both arterioles and capil-laries take part in substance exchange (2) and cerebrovascularresistance (3, 4). Specifically, modified smooth muscle cells calledpericytes are attached to capillaries and can regulate CBF at thecapillary level (3, 5, 6). However, this regulation is not completelyunderstood. Retinal pericytes are constricted by ATP and dilatedby neurotransmitters in vitro (5), and they constrict in vivo fol-lowing stroke (7). In response to light stimulation, retinal capil-laries actively dilate and regulate blood flow independent ofarterioles (8). Furthermore, glial Ca2+ signaling regulates capil-lary, but not arteriole, blood flow in both the retina and the ce-rebral cortex (8, 9). Nevertheless, capillary pericytes have beensuggested to not be contractile, and that the regulation of CBF inthe CNS is only mediated by smooth muscle cells in penetratingarterioles (p.a.s) and capillaries, but not by pericytes on capillaries(10–12). This controversy may be more apparent than real be-cause it depends on how a pericyte and capillary are definedrather than the role of brain capillaries in cerebrovascular control.

We have chosen to analyze the change in brain capillaries basedon the branching order from the p.a. (13). Using this unbiasedmethodology, our study may contribute to understanding the con-tribution of capillaries and pericytes to cerebrovascular control andthe interplay between capillaries and arterioles.All capillaries have pericyte coverage (14), and pericytes are

almost completely covered by astrocyte end-feet (15), which raisesthe possibility that soluble signal molecules released into the mi-croenvironment by astrocytes are sensed by specialized surfacereceptors on pericytes. ATP is the main transmitter by which as-trocytes communicate with neighboring astrocytes (16), as well asan important paracrine transmitter in signaling to neurons (17)and possibly pericytes (18). Therefore, an important part of thisstudy was an examination of the effect of ATP on brain capillarypericytes in vivo.The current study used in vivo two-photon microcopy of a

transgenic mouse model with fluorescent pericytes. The activity-dependent increase in synaptic activity was examined to de-termine whether capillaries of all branching orders are dilated orconstricted, or only capillaries close to the p.a., and whethercapillaries exhibit conducted vascular responses (CVRs) similarto pial arterioles.Our study supports the notion that pericytes play active roles

in neurovascular coupling. Furthermore, the results suggest thatboth arterioles and capillaries contribute to cerebrovascular con-trol during physiological stimulation, and that spatially restrictedCVRs may contribute to regulating the flow in brain capillariesand the spatiotemporal characteristics of functional neuroimagingsignals.

Significance

Pericytes are located at the outside wall of capillaries. How-ever, whether and how pericytes are involved in the regulationof blood flow in brain capillaries is still debated. We report thatcapillary vascular responses are mostly initiated and peak atnear-arteriole capillaries. These vascular responses are con-ducted along capillaries at a speed of 5–20 μm/s. Conductedvascular responses in brain capillaries appear to involve peri-cytes, the mural cells of microvessels, and may be a novelmodulator of vascular function in the brain.

Author contributions: C.C., B.B., and M.J.L. designed research; C.C., J.C.F., S.H.J., M.L., andS.A.Z. performed research; C.C. and B.G. contributed new reagents/analytic tools; C.C. andB.O.H. analyzed data; and C.C., B.O.H., and M.J.L. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Published under the PNAS license.1To whom correspondence should be addressed. Email: mlauritz@sund.ku.dk.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1707702115/-/DCSupplemental.

Published online June 4, 2018.

E5796–E5804 | PNAS | vol. 115 | no. 25 www.pnas.org/cgi/doi/10.1073/pnas.1707702115

Dow

nloa

ded

by g

uest

on

Oct

ober

9, 2

020

21%

55%

24%

penetratingarteriole1st ordercapillarieshigher ordercapillaries Distance from p.a. (μm) Distance from p.a. (μm)

Upstream

Downstream

CVR speed

Late

ncy

(s)

Latency

Dia

met

er C

hang

e (%

)

Amplitude

p.a. 1stcap 2ndcap ≥3rdcap0

10

20

30

Pea

k am

plitu

de (%

)

29 29 23 140

5

10

15

20

Spe

ed (μ

m/s

)

n.s.

16 17

***

p.a. 1stcap 2ndcap ≥3rdcap0

5

10

15

20

Dia

met

er (μ

m)

42 42 25 14

p.a. 1stcap 2ndcap ≥3rdcap

*

0

5

10

15

Late

ncy

(s)

29 29 23 8

*

*

Whisker pad stimulation

0 10 20 30 40 50 60

0

20

40

60

80

100

Time (s)

p.a.1st cap2nd cap3rd cap

Dia

met

er (%

of p

eak)

20 40 6020304050

No. of frames

Dia

met

er(p

ixel

)

0 80 100

p.a.

1st cap

2nd cap 3rd cap

0 50 100 1500

5

10

15

20

0 50 100 15010

20

30

40

50

60

70

p.a.1st cap2nd cap3rd cap

Initiation of dilation

*

** *

A

E F

I J

H

K

B D

C

G

Fig. 1. Functional vessel dilation in the mouse barrel cortex. (A) A two-photon image of the barrel cortex of a NG2-DsRed mouse at ∼150 μm depth. The p.a.sbranch out a capillary horizontally (first order). Further branches are defined as second- and third-order capillaries. Pericytes are labeled with a red fluorophore(NG2-DsRed) and the vessel lumen with FITC-dextran (green). ROIs are placed across the vessel to allow measurement of the vessel diameter (colored bars). (Scalebar: 10 μm.) (B) Vessel diameters at different orders of capillaries. p.a., 15.09 ± 4.15 μm; 1st cap (first-order capillaries), 7.18 ± 1.93 μm; 2nd cap (second-ordercapillaries), 6.25 ± 2.43 μm; 3rd cap (third-order capillaries), 7.63 ± 2.47 μm. The p.a. diameter is significantly larger than all orders of capillaries. ***P < 0.001, one-way ANOVA with post hoc test. (C) Example trace of fluorescent intensity over time at the blue ROI in A is shown as the gray image, and the two red curvesindicate the vessel wall (Upper). The distance of the two red curves is calculated as the time course of vessel diameter (Lower). (D) The normalized diameter changeover time at different orders of capillaries in response to whisker-pad stimulation. The short vertical bar is where the curve reaches 50% peak, which is defined asresponse onset. (E) Distribution of the locations where the functional dilation initiated (n = 29 locations). (F) Multiple ROIs at the p.a. and first-, second-, and third-order capillaries are marked as red, blue, green, and yellow, respectively. (Scale bar: 10 μm.) (G) In this mouse experiment, the half-maximal dilation latency of eachROI is plotted with corresponding colors on the left along the geographic distance from the p.a. Dashed lines show the linear fit of the conducted dilation. (H) Themaximal dilation amplitude is plotted with corresponding colors on the left along the geographic distance from the p.a. (I) Eighteen out of 29 imaged vasculaturesexhibited conducted functional dilation, with an upstream conductive speed of 12.65 ± 0.96 μm/s and downstream conductive speed of 12.83 ± 0.64 μm/s. Nosignificant difference was found between upstream and downstream conductive speeds. n.s., not significant; P > 0.05, unpaired t test. (J) Time to 50% maximaldilation was significantly longer in the third-order capillaries than the p.a. and first-order capillaries. The second-order capillaries dilated significantly slower thanthe first-order capillaries. *P < 0.05, one-way ANOVA with post hoc test. (K) Maximal dilation amplitude in different order capillaries. First- and second-ordercapillaries exhibited significantly larger responses than other locations. *P < 0.05, one-way ANOVA with post hoc test. All error bars represent SEM.

Cai et al. PNAS | vol. 115 | no. 25 | E5797

NEU

ROSC

IENCE

PNASPL

US

Dow

nloa

ded

by g

uest

on

Oct

ober

9, 2

020

ResultsCVRs Initiate at Capillaries or p.a.s. We used in vivo two-photonmicroscopy to image the vasculature in the whisker-barrel cortexof anesthetized mice expressing DsRed in pericytes under con-trol of the NG2 promotor. FITC-dextran was used to label theblood plasma (shown as green in Fig. 1A). The p.a.s were iden-tified unequivocally in vivo by tracing their connections back tothe pial arterioles and by the clear continuous rings of smoothmuscle around them. Only p.a.s with a longitudinal axis per-pendicular to the x–y plane were used for data analysis. Capil-laries were identified as microvessels branching off from the p.a.with a longitudinal axis parallel to the x–y plane. This geometricarrangement was necessary for reliable measurement of changesin the arteriolar and capillary diameter. Pericytes were identifiedas NG2-positive mural cells on capillaries branching off from thearteriole. Pericyte cell bodies were spatially separated from thep.a. and each other, and individual pericytes were identified byprocesses extending longitudinally along capillaries (Fig. 1A).Based on z-stacks of the cortex, we segmented the blood

vessels by branching order, 0 being the p.a., 1 being the first-order capillary branching off the arteriole, and so on (Fig. 1A).In the resting state, the diameters of the p.a. and first-, second-,and third-order capillaries were 15.09 ± 0.10 μm, 7.18 ± 0.04 μm,6.25 ± 0.10 μm, and 6.75 ± 0.28 μm, respectively. The p.a. wassignificantly wider than the capillaries, but the capillary diameterwas similar among the first three orders of capillaries (Fig. 1B).Reportedly, first-order capillaries dilate first in response to

somatosensory stimulation and the time to vasodilation in first-order capillaries commonly precedes dilation in the p.a. (3). Inthe present study, the time resolution did not allow us to assessdifferences in the time of onset of the stimulation-induced va-sodilation, but as a proxy we used the latency from stimulus onsetto 50% maximal dilation (Fig. 1D and Movie S1). Out of 29preparations, stimulation-evoked dilation was achieved first infirst-order capillaries in 55% of experiments, whereas dilationwas achieved first in the p.a. in 21% and in second- or third-order capillaries in 24% of experiments (Fig. 1E). Next, weevaluated whether a pattern exists in the development of capil-lary dilation, that is, whether dilation occurs first at a particularpoint and whether the reaction spreads according to the branchingorder of the capillaries. For this purpose, multiple regions of in-terest (ROIs) rectangles with the long side perpendicular to thevessel wall were drawn as indicated by the color coding in Fig. 1F.The half-maximal latency and maximal vascular dilation for eachROI was plotted as a function of the geographic distance alongthe vasculature from the p.a. using the same color coding as thesquares representing ROIs (Fig. 1 G and H). Fig. 1G shows the x–yplane of one mouse. Dilation initiated at the first-order capillary,and the dilation propagated to the p.a. and second and thirdcapillaries in a linear fashion. The first- and second-order capil-laries demonstrated the strongest dilation (Fig. 1H). The timesequence of vascular reactions could be reported for 18 of the29 experiments; in 11 experiments the baseline stability was sub-optimal. In the experiments with a stable enough baseline, theaveraged upstream conductive speed was 12.65 ± 0.96 μm/s andaveraged downstream conductive speed was 12.83 ± 0.64 μm/s(Fig. 1I). The vascular dilation spread with the same velocityupstream and downstream (P = 0.67). Dilation was significantlyslower in the second-order and higher capillaries than in the first-order capillaries and the p.a. (Fig. 1J), whereas changes in di-ameter were significantly greater in the first- and second-ordercapillaries than in the p.a. and higher-order capillaries (Fig. 1K).To exclude the possibility that the conducted responses were

affected by focus drift, we carried out hyperstack imaging (con-tinuous and repetitive recordings of z-stack images) duringwhisker-pad stimulation. Images were flattened to produce atime-series movie by maximum intensity projections for each image

stack. This procedure confirmed CVRs in the p.a.s and capillariesupon whisker stimulation, both upstream and downstream, infive of five experiments (Supporting Information, Fig. S1, andMovie S2).

CVRs Induced by Local ATP Injection. Purinergic signaling may affectthe neurovascular unit in pathological states, such as during ce-rebral ischemia when ATP is released in high concentrations(19–22). ATP constricts retinal pericytes and capillaries in vitro,which is of interest because brief periods of ischemia lead to theno-reflow phenomenon and a reduction in the caliber of smallvessels (7, 23). We examined the effect of purinergic receptoractivation on pericytes and capillaries in vivo by local injection ofATP into the barrel cortex of NG2-DsRed mice. Guided by thetwo-photon microscope, a glass micropipette was inserted intothe cortex and advanced to close proximity of the p.a. and thefirst few orders of capillaries. A mixture of 10 μM Alexa 594 (redcolor in the glass micropipette) and 1 mM ATP was puffed fromthe micropipette by air pressure (Fig. 2 A and B). ATP puffingevoked capillary dilation, followed by constriction (Fig. 2B andMovie S3). Fifteen rectangular ROIs were studied at different-order capillaries (Fig. 2C) and the normalized diameter changewas plotted over time for each ROI (Fig. 2D). Amplitudes ofdilation or constriction were defined as positive or negative am-plitudes at maximal vascular response. The latency of dilationsand constrictions were reported as the time to half positive ornegative maximum after puffing onset (Fig. 2E). The four vari-ables were plotted as a function of the geographic distance alongthe vasculature from the p.a. (Fig. 2 F–I). The same color codingwas used for the squares representing ROIs. The branching pointof the first- to second-order capillary exhibited the strongest andearliest dilation and constriction, whereas the third-order capillaryhad a very small change in diameter (Fig. 2 F and G). A signifi-cantly higher amplitude of both vasodilation and vasoconstrictionwas found at the first- and second-order capillaries, whereas thediameters of higher-order capillaries were almost unaltered (n =7; Fig. 3 A and B). The latencies of vasodilation and vasocon-striction at third-order and higher capillaries were significantlylonger than at lower-order capillaries (Fig. 3 C and D). No sig-nificant difference was found for the mean distance from the pi-pette tip to the different-order capillaries, which excludes aninfluence of distance to pipette tips on the conducted responses(Fig. 3E). Furthermore, traces from individual mice indicated nocorrelation between pipette distance and the response properties(i.e., latency and amplitude) (Fig. S2).ATP-puffing-induced dilation and constriction demonstrated

linear or near-linear conduction in the upstream and down-stream direction (Fig. 2 H and I). The conductive speed of di-lation to upstream and downstream vessels was 11.47 ± 3.37 μm/sand 14.78 ± 3.85 μm/s, respectively, whereas the conductivespeed of constriction to upstream and downstream vessels was6.54 ± 1.05 μm/s and 6.55 ± 1.22 μm/s, respectively, that is, slowerthan for dilation (Fig. 3F). Hyperstack imaging during ATP puffingconfirmed conductive responses for both ATP-induced dilation andconstriction in five of five experiments (Supporting Information, Fig.S3, and Movie S4). The faster conductive speed for downstreamdilation suggests that ATP-induced conducted vascular dilation andconstriction are modulated by different mechanisms.

ATP-Puffing-Induced Constriction, but Not Dilation, Depended onPurinergic Type 2 Receptors. To probe the mechanism of ATP-puffing-induced constriction, 0.5 mM of the P2 receptor antag-onist pyridoxalphosphate-6-azophenyl-2′,4′-disulfonic acid (PPADS)was applied topically to the exposed cortex at least 2 h before ATPpuffing. Preconditioning with PPADS preserved vessel dilation butprofoundly attenuated vessel constriction (Fig. 4 A and B). Thesensitivity to PPADS was particularly pronounced in first- andsecond-order capillaries (Fig. 4 C and D). This indicates that the

E5798 | www.pnas.org/cgi/doi/10.1073/pnas.1707702115 Cai et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

9, 2

020

ATP-induced constriction observed in the absence of inhibitor atfirst- and second-order capillaries, and to a lesser extent at the p.a.and third-order and higher capillaries (Fig. 3), was due to P2 puri-nergic receptor activation.

Activation of both P2X and P2Y Receptors Leads to Similar VesselResponses. In vitro studies of arteries and arterioles have shownthat the activation of P2Y receptors on smooth muscle cells leadsto vessel constriction, whereas the activation of P2X receptors onarteriolar endothelial cells (ECs) leads to vessel dilation (24–27).To test whether the same mechanisms contribute to brain cap-illary control in vivo, we investigated the vessel responses elicited

by both P2X and P2Y receptor agonists. P2X receptor agonistαβATP (αβ-methylene-ATP) and P2Y receptor agonist UTPwere administered (1 mM each) by puffing in close proximity tothe p.a. and the first few order capillaries. As adenosine hydro-lyzed from ATP is a potent vasodilator (28), a more stable ATPanalog, ATPγS, was used at a concentration of 1 mM for micro-pipette puffing experiments. Finally, control experiments wereperformed by puffing 10 μM Alexa 594 only to rule out the effectof puffing itself.We compared the effect of the compounds on the first-order

capillary responses because these capillaries had the most robustand profound responses upon ATP puffing (Fig. 3; see also Fig.

Latency of dilation

Puf

f

10% ofØ change

p.a.

1st cap

2nd cap

3rd cap

0 20 40 60 80 100 120−50−40−30−20−10

0102030

Time (s)

Dia

met

er c

hang

e (%

)

Amplitude of dilation

Amplitude of constriction

Latency of constriction0 50 100 150

0

10

20

30

Distance from p.a.(μm)

Dia

met

er c

hang

e (%

)

Amplitude of dilation

0 50 100 150

−60

−40

−20

0Distance from p.a.(μm)

Dia

met

er c

hang

e (%

)

Amplitude of constriction

0 50 100 1500

2

4

6

8

Distance from p.a.(μm)

Late

ncy

(s)

Latency of dilation

0 50 100 15015

20

25

30

35

Distance from p.a.(μm)

Late

ncy

(s)

Latency of constriction

64.7 s 83.1 s 119.4 s

0.6 s 45.7 s 52.3 s

10 s

ATP

Puf

f

A B

C F G

H ID

E

Fig. 2. ATP puffing by micropipette induces vessel dilation, followed by constriction. (A) Diagram of the in vivo experimental setup. The puffing micropipetteis placed in proximity of the near-arteriole site. The micropipette contains a mixture of 10 μM Alexa 594 (red color in glass micropipette) and 1 mM ATP. (B)Video snapshots of the time course of puffing with 1 mM ATP from the micropipette. Vessel dilation precedes constriction. Dashed lines indicate the vesselcontours at the resting state. (Scale bars: 10 μm.) (C) Multiple uniquely colored ROIs are placed along the vasculature to measure the vessel diameter. (Scalebar: 10 μm.) (D) Normalized diameter change is plotted over time for each ROI. The ROIs and trace color are coded identically. (E) Amplitudes of dilation orconstriction are defined as positive or negative amplitudes at the maximal vascular response. The latencies of dilation and constriction are reported as time tohalf positive or negative maximum after puffing onset. (F–I) In this mouse experiment, the distribution of all ROIs in C and D with amplitude of dilation (F),amplitude of constriction (G), latency of dilation (H), and latency of constriction (I) over the geographic distance from the p.a. along the vasculature. Thedashed lines represent the linear fit of the upstream and downstream conductive responses.

Cai et al. PNAS | vol. 115 | no. 25 | E5799

NEU

ROSC

IENCE

PNASPL

US

Dow

nloa

ded

by g

uest

on

Oct

ober

9, 2

020

S4). Both P2X and P2Y receptor agonists induced dilation of thefirst-order capillaries, followed by constriction. Although theamplitudes of dilation were the same for all compounds (Fig.5A), control experiments using only Alexa 594 showed only weakdilation in one of the five experiments. The amplitude of vaso-constriction was significantly larger with ATPγS than ATP (Fig.5B), which could be explained by ATPγS inducing vasocon-striction via activation of both the P2X and P2Y receptors. Thedilator effect is less pronounced because ATPγS is more stablethan ATP, with the formation of less adenosine. In control ex-periments, vessel constriction was rarely observed. Both vasodi-lation and vasoconstriction took longer to develop when evokedby ATP than for the compounds that were more stable (Fig. 5 Cand D). This may be explained by the vasodilator adenosinehydrolyzed from ATP prolonging the latency to maximal vaso-constriction. A comparison of upstream and downstream con-ductive speeds of both vasodilation and constriction with the fourcompounds found no significant difference (Fig. 5E), suggesting

that the four compounds exert the conducted vasoresponses viaa similar mechanism.

Instant and Severe Pericyte Constriction After Ischemia. To study theinstant response of pericytes in ischemic stroke in vivo, cardiacarrest was induced by i.v. injection of 0.05 mL pentobarbital. Animage stack covering the whole near-arteriole region was recordedbefore cardiac arrest, and another image stack in the same regionwas recorded 5 min after cardiac arrest. Each image stack wasthen projected onto one image by average intensity projection.The p.a. and first- and second-order capillaries exhibited severeconstriction, in some cases with red blood cells clogged inside, butonly adjacent to pericyte cell bodies. The third-order or highercapillaries exhibited less constriction, though they also had muralpericytes (Fig. 6 A and C). Capillaries with visible pericyte cellbodies exhibited more constriction than capillaries devoid ofpericyte cell bodies (Fig. 6C).Next, we evaluated whether purinergic receptors were in-

volved in the ischemia-induced capillary constriction by peri-cytes. For this purpose, we superfused the exposed mouse cortexwith 0.5 mM PPADS for at least 2 h before cardiac arrest. Thismitigated constriction of the p.a. and first- and second-ordercapillaries (Fig. 6 B and C). These results are consistent with priorstudies of brain slices and postmortem studies suggesting thatpericytes constrict in ischemia (3, 7) and that preconditioning theanimals with PPADS helps in the recovery from experimentalstroke (29, 30). We conclude that pericytes at the first severalorders of capillaries constrict severely after ischemia in vivo andthat blocking purinergic receptors mitigates the constriction ofboth arterioles and capillaries.

DiscussionUnderstanding neurovascular signaling in response to neuronalor astrocytic activity is crucial to understanding how brain pro-cesses are supplied with energy and functional neuroimagingsignals are generated. Our results demonstrate that first- andsecond-order capillaries initiate functional dilation more oftenthan the p.a. and higher-order capillaries. In addition, local anddirect administration of ATP induces vessel dilation, followed byconstriction at the first several orders of capillaries. Functionaldilation and ATP-puffing-induced dilation and constriction areinitiated mostly at the first- or second-order capillaries, and CVRsdevelop both upstream and downstream. However, the velocityof conducted vasodilation is faster than for conducted vasocon-striction. Furthermore, minutes after cerebral ischemia, pericytesat near-arteriole sites constrict via a P2 receptor-dependentmechanism.The contribution of pericytes to CBF regulation has been

controversial (13). Some in vivo studies suggest a role of peri-cytes in the regulation of capillary blood flow (3, 6), whereasothers have indicated that flow control was detectable only inarterioles, but not in capillaries, and that vascular smooth musclecells, but not pericytes, contribute to the regulation of CBF re-sponses (10–12). However, most of those studies have suggestedthat pericytes close to the p.a. are contractile during normalbrain activity, and pericytes on first- and second-order capillarieshave hybrid features of both smooth muscle cells and capillarypericytes (31). In our studies, pericytes were identified by two-photon microscopy as red-fluorescent cells on the capillary wallin NG2-DsRed mice. Nevertheless, we describe the changes incapillary function according to the branching orders from the p.a.and involvement of pericytes in this context. Our results strengthenthe importance of using a defined vessel geometry with respect tothe cortical surface to reliably assess small changes in the diame-ters of capillaries.Similar to other studies (3, 10), our data suggest a key role of

capillaries close to the p.a. in local blood flow control. As a powerfultool for studying CBF (32), the two-photon imaging microscope was

A B

C Dp.a. 1st 2nd ≥3rd

0

5

10

15

20

25

Dia

met

er c

hang

e (%

) Amplitude of dilation

**

*

6 7 7 4

*0

-10

-20

-30

-40

Amplitude of constriction

Dia

met

er c

hang

e (%

)

0

5

10

15

Latency of dilation

Tim

e (s

)

0

10

20

30

40

50Latency of constriction

Tim

e (s

)*

6 7 7 4

**

6 7 7 4

*

F

0

5

10

15

20

CVR speed

Spe

ed (μ

m/s

)

**

5 5 65

6 7 7 4

*

0

20

40

60

80

100

Distance to pipette tip

Dis

tanc

e (μ

m)

n.s.

E

6 7 7 4

dilation

upstream

dilation

downstream

constrictio

n

upstream

constrictio

n

downstream

p.a. 1st 2nd ≥3rd

p.a. 1st 2nd ≥3rd

p.a. 1st 2nd ≥3rd

p.a. 1st 2nd ≥3rd

Fig. 3. ATP-puffing-induced dilation and constriction vary in amplitude andlatency at different-order capillaries. (A) The amplitudes of dilation and (B)constriction are significantly higher at first- and second-order capillariescompared with other order capillaries. (C) The latencies of dilation and (D)constriction are significantly longer for third-order and higher capillariesthan other order capillaries. *P < 0.05, one-way ANOVA with post hoc test.(E) The mean distance from the pipette tip to the vessels of the differentbranch orders. n.s., not significant; P > 0.05, one-way ANOVA with post hoctest. (F) Comparison of upstream and downstream conductive speeds ofATP-puffing-induced dilation and constriction. *P < 0.05, one-way ANOVAwith post hoc test. All error bars represent SEM.

E5800 | www.pnas.org/cgi/doi/10.1073/pnas.1707702115 Cai et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

9, 2

020

used to focus on one horizontal plane at a depth of 100–200 μm toobtain good image quality. This horizontal plane most commonlyincluded one p.a. and the associated first-, second-, or third-ordercapillaries. Our results showed that capillary dilation as a responseto increased synaptic activity (i.e., the neurovascular coupling re-sponse) is initiated in most cases in first-order capillaries. This isconsistent with earlier results indicating active relaxation of peri-cytes before relaxation of arteriolar smooth muscle cells (3), andwith recent studies showing that smooth muscle actin is present inpericytes at near-arteriole capillaries (10, 33). In addition, pericytesat near-arteriole capillaries may have denser smooth muscle actin(12, 34), as the first- and second-order capillaries react with theearliest and most profound dilation.ATP puffing induced dilation and constriction in the first few

orders of capillaries. ATP puffing onto higher than second-ordercapillaries induced almost no changes in diameter. The obser-vations indicate that pericytes have different sensitivities topurinergic stimulation, in accordance with previous studies (5).Dilation induced by ATP puffing has been suggested to be me-diated by the activation of P2X receptors on arteriolar ECs (24).However, a recent study proposed that ATP may also act onastrocytic P2X1 receptors to evoke the release of PGE2, whichrelaxes pericytes (9). Other studies have shown that the ATPanalog ATPγS has its own pharmacological profile; for example,it may in fact be hydrolyzed to adenosine (35), which may beanother explanation for ATP-puffing-induced vasodilation.ATP-puffing-induced vessel constriction was profoundly at-

tenuated by the purinergic type 2 receptor antagonist PPADS,indicating the involvement of purinergic type 2 receptors. Similarto our study, ATP puffing in brain slices was previously shown toincrease cytosolic Ca2+ in glial cells, followed by adjacent vaso-constriction, which was abolished by preincubation with P2Y1receptor blocker (36). Our in vitro studies with pericytes in

monoculture (Supporting Information, Fig. S5, and Movie S5)confirmed that ATP constricted capillary pericytes in a mannerdependent on P2 receptor activation and increased cytosolicCa2+. This study demonstrates that intracellular Ca2+ increasesin pericytes may be the mechanism underlying pericyte con-traction in response to ATP. ATP applied to cerebral arteriolesin vitro produced a biphasic vessel response, constriction fol-lowed by dilation (24), which is the opposite of what we observedin capillaries. The data suggest that the effect of the purinergicsignaling cascade in capillaries is different from the effect inarterioles (9).In arteries and arterioles, CVRs are primarily characterized by

fast (1–3 mm/s) and far-reaching electrical conduction along well-coupled endothelium and into smooth muscle (37, 38). UponG-coupled receptor stimulation, a secondary slow and spatiallylimited Ca2+ wave spreads along the endothelium (∼100 μm/s),giving rise to nitric oxide and prostaglandin production (Fig. 7 Aand B). Inhibition of the electrical component has demonstratedthat the speed of the slow, diffusion-based CVR is ∼20 μm/s,similar to the slow speed of propagated vasodilation observedin the present study (39). Furthermore, hyperpolarizing pulsespropagate along the ECs in capillaries, with a conductive speed100 times faster than the diffusion-mediated CVR (37). In con-trast, the vascular relaxation times are the same for the two typesof conducted responses. It is possible that the final common pathof both types of vascular responses may involve axon boutonsdumping potassium concurrent with rapid spiking (40), but this willneed to be addressed in more detail in future studies.Although the underlying mechanism remains unclear, a diffusion-

based conduction of vasomotor responses emanating from first- andsecond-order capillaries can be envisioned, for example paracrinesignaling along astrocytic end-feet or intracellular diffusion acrossgap junctions connecting ECs and/or pericytes (Fig. 7C). However,

pa 1stcap 2ndcap 3rdcap0

5

10

15

20

25

30Amplitude of dilation

Dia

met

er c

hang

e (%

) ATPATP (preconditioned w/ PPADS)

pa 1stcap 2ndcap 3rdcap0

-10

-20

-30

-40

Amplitude of constriction

Dia

met

er c

hang

e (%

)ATPATP (preconditioned w/ PPADS)

**

6 7 7 48 8 6 5

6 7 7 48 8 65

0 20 40 60 80 100 120-10

0

10

20

Dia

met

er c

hang

e (%

)

Time (s)

A B

C D

Fig. 4. ATP-puffing-induced constriction is mediated by purinergic type 2 receptors. (A) After topical application of 0.5 mM PPADS, the puffing pipette isplaced in proximity of a near-arteriole site. Red, blue, green, and yellow ROIs are placed at the p.a. and first-, second-, and third-order capillaries to measurethe diameter, respectively. (Scale bar: 10 μm.) (B) Time course of the diameter change in each ROI indicated in A. Preconditioning with 0.5 mM PPADS andpuffing with 1 mM ATP induces dilation but profoundly attenuates constriction. (C) The amplitude of dilation is not significantly different with and withoutthe application of PPADS among all orders of capillaries. n.s., not significant; P > 0.05, unpaired t test. (D) The amplitude of constriction is significantlydifferent with and without preconditioning with PPADS on the first- and second-order capillaries. *P < 0.05, unpaired t test. All error bars represent SEM.

Cai et al. PNAS | vol. 115 | no. 25 | E5801

NEU

ROSC

IENCE

PNASPL

US

Dow

nloa

ded

by g

uest

on

Oct

ober

9, 2

020

documentation of gap junction coupling between the ECs of first-and second-order capillaries in mouse cortex is lacking. In addi-tion, in contrast to arterial endothelium, capillary ECs do notharbor SKCa/IKCa channels (37), which are thought to underlie theinitiation of fast electrical CVRs (41). This may also explain whymicroapplication of ATP to pial arteries and p.a.s in vitro hasbeen observed to result in constriction followed by endothelium-dependent propagated vasodilation (25, 26).During ischemia, cerebral pericytes constrict and stop the

blood flow in capillaries within a few minutes. The constriction ispronounced at the p.a. and first- and second-order capillaries,whereas the diameters of higher capillaries remained constant.This supports and modifies the notion that pericytes contributeto the long-lasting decrease in capillary blood flow after cerebralischemia (3, 7, 42). The most ischemia-sensitive region of thevasculature is first-order capillaries at the near-arteriole site.Preconditioning with the P2 receptor antagonist PPADS allevi-ated pericyte constriction after ischemia, which may be related tothe blockade of purinergic type 2 receptors, but its effect inpreventing pericyte constriction in ischemia is equally likely toreflect that it blocks the reversed mode of Na+/Ca2+ exchange,

a well-known route for Ca2+ overload in ischemia (43). Moreover,PPADS blocks ecto-ATPases (44), thereby promoting pericyteconstriction after ischemia. Other studies have reported that thatmodulation of purinergic receptors promotes animal recoveryfrom stroke in vivo (29, 30), but our studies show that pericyteson first-order capillaries can be rescued by blocking purinergicreceptors before cerebral ischemia.

A

B

Cp.a. 1st cap 2nd cap 3rd cap ≥4th cap

−100

−80

−60

−40

−20

0

pericyte soma; (-) PPADSno pericyte soma; (-) PPADSpericyte soma; (+) PPADSno pericyte soma; (+) PPADS

* ****

*

Dia

met

er c

hang

e (%

)

15 15 12 12 67 14 13 106 5 5 5 54 5 5 4

*

Prec

ondi

tione

d w

ith P

PAD

SW

ithou

t PPA

DS

n/a n/a

Before CA 5 min after CA

Before CA 5 min after CA

−120

Fig. 6. Ischemia leads to severe constriction of capillaries at the near-arteriole site and preconditioning of P2 receptor blockers mitigates con-striction of capillaries. (A) Image stacks (1-μm step size, average intensityprojection) of the vasculature, including the p.a. and first few orders ofcapillaries. Five minutes after ischemia by cardiac arrest (CA), severe con-striction was observed at the p.a. and first- and second-order capillaries, butthird-order and higher capillaries were moderately constricted. Dashed linesindicate the vessel contours of first-order capillaries before cardiac arrest.(Scale bars: 20 μm.) (B) Preconditioning with 0.5 mM PPADS for 2 h rescuedsevere constriction of the p.a. and first-order capillary 5 min after CA.Dashed lines indicate the vessel contours of first-order capillaries before CA.(Scale bars: 20 μm.) (C) The most severe constrictions at the first- and second-order capillaries colocalized with pericytes. The third-order and highercapillaries exhibited moderate constriction. Preconditioning with PPADSmitigated vasoconstriction at the p.a. and first- and second-order capillaries.For the p.a., an unpaired t test was used. n/a, not available; ***P < 0.001. Forthe other order capillaries, one-way ANOVA with post hoc test was used(*P < 0.05). All error bars represent SEM.

A B

C D

0

5

10

15

20

25Amplitude of dilation

Dia

met

er c

hang

e (%

) 0

-10

-20

-30

-40

-50

-60

Amplitude of constriction

Dia

met

er c

hang

e (%

)

0

2

4

6

8

Tim

e (s

)

0

5

10

15

20

25

30Ti

me

(s)

Latency of constriction*

n/a7 6 5 5

E

***7 6 5 5

5

Dilation

upstream

0

10

20

30CVR speed

Spe

ed (μ

m/s

) ATPαβATPUTPATPγS

Constriction

upstreamDilation

downstream

Constriction

downstream

5 5 5 65 3 5 54 4 4 45 5 54

n.s.

***

7 6 5 5 5

*

n.s. n.s.n.s.

**

*Latency of dilation

7 6 5 5 n/a

Control

Fig. 5. Vessel responses of first-order capillaries to puffing with ATP, P2X,P2Y receptor agonists, ATPγS, and red dye. (A) Comparison of differentpuffing compounds with amplitude of dilation, (B) amplitude of constric-tion, (C) latency of dilation, (D) latency of constriction, and (E) conductivespeed. The compounds are 1 mM ATP, 1 mM P2X receptor agonist (αβATP),1 mM P2Y receptor agonist (UTP), 1 mM ATPγS, and 10 μM Alexa 594 ascontrol. n/a, not available; *P < 0.05, ***P < 0.001, one-way ANOVA withpost hoc test. Note that the latency of the control experiment is marked asnot available for both dilation (C) and constriction (D). This is due to thesmall responses upon control puffing and the suboptimal measurements oflatency. All error bars represent SEM. n.s., not significant.

E5802 | www.pnas.org/cgi/doi/10.1073/pnas.1707702115 Cai et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

9, 2

020

Materials and MethodsAnimal Handling. All procedures involving animals were approved by theDanish National Ethics Committee according to the guidelines set forth in theEuropean Council’s Convention for the Protection of Vertebrate Animals Usedfor Experimental and Other Scientific Purposes and were in compliance withthe ARRIVE guidelines. Forty NG2-DsRed mice [Tg(Cspg4-DsRed.T1)1Akik/J;Jackson Laboratory] of both sexes were used at 4–7 mo of age. The tracheawas cannulated for mechanical ventilation (Minivent type 845; HarvardApparatus) and catheters were placed into the left femoral artery andvein for the infusion of substances and to monitor blood pressure and bloodgases. To ensure that the animals were kept under physiological conditions,we continuously monitored end-expiratory CO2 (Capnograph type 340; HarvardApparatus) and blood pressure (pressure monitor BP-1; World Precision In-struments) and assessed blood gases in arterial blood samples twice duringeach experiment (pO2, 95–110 mmHg; pCO2, 35–40 mmHg; pH, 7.35–7.45)using an ABL 700Series radiometer. Body temperature was maintained at37 °C using a rectal temperature probe and heating blanket (TC-1000 Temper-ature Controller; CWE).

The experimental setup involved gluing the skull to a metal plate withcyanoacrylate gel (Loctite Adhesives). A 4-mm-diameter craniotomy wasdrilled, centered 0.5mmbehind and 3mm to the right of the bregma over thesensory barrel cortex region. After removing the dura, the preparation wascovered with 0.75% agarose gel (type III-A, low EEO; Sigma-Aldrich),moistened with artificial cerebrospinal fluid (aCSF; NaCl 120 mM, KCl2.8 mM, NaHCO3 22 mM, CaCl2 1.45 mM, Na2HPO4 1 mM, MgCl2 0.876 mM,and glucose 2.55 mM; pH 7.4), and kept at 37 °C. For imaging experiments,part of the craniotomy was covered with a glass coverslip that permitted theinsertion of electrodes and pharmacological interventions.

The mice were anesthetized by i.p. injection of a mixture of ketamine(60 mg/kg) and xylazine (10 mg/kg) and administered supplemental doses(30 mg/kg) of ketamine every 20 min. Upon completion of all surgical pro-cedures, the anesthesia was switched to continuous i.v. infusion withα-chloralose (33% wt/vol; 0.01 mL/10 g/h). At the end of the experimentalprotocol, the mice were killed by i.v. injection of 0.05 mL pentobarbitalfollowed by cervical dislocation.

Whisker-Pad Stimulation. The mouse sensory barrel cortex was activated bystimulation of the contralateral ramus infraorbitalis of the trigeminal nerveusing a set of custom-made bipolar electrodes inserted percutaneously. Thecathode was positioned relative to the hiatus infraorbitalis (IO), and theanode was inserted into the masticatory muscles (45). Thalamocortical IOstimulation was performed at an intensity of 1.5 mA (ISO-flex; A.M.P.I.) for1 ms in trains of 20 s at 2 Hz.

Micropipette Puffing. The glass micropipettes for puffing were produced by apipette puller (P-97; Sutter Instrument) with a resistance of 3–3.5 MΩ. Thepipette was loaded with a mixture of 10 μM Alexa 594 and active substancesto visualize the pipette under the two-photon microscope. Guided by thetwo-photon microscopy, the pipette was inserted into the cortex and thevasculature approached 100–200 μm below the surface. The distance be-tween the pipette tip and vasculature was 30–50 μm (Fig. 2A). Substanceswere puffed for ∼200 ms using an air pressure of ∼15 psi in the pipette. Thepipette tip was placed randomly near the p.a. or first- or second-ordercapillaries and the dye spread very quickly (∼160 μm/s). Within one to twoframe acquisition times the “red cloud” covered both the arteriole and thecapillaries, and the background returned to normal ∼20 s after puffing. Thevasoresponses were not affected by the concentration of ATP at differentdistances (Fig. S2).

Two-Photon Imaging. FITC-dextran (2% wt/vol, molecular weight 70,000,50 μL; Sigma-Aldrich) was administered into the femoral vein to label the bloodplasma. Experiments were performed using a commercial two-photon mi-croscope (Femto3D-RC; Femtonics Ltd.) and a 25 × 1.0 N.A. water-immersionobjective. The excitation wavelength was set to 900 nm. The emitted lightwas filtered to collect red and green light from DsRed (pericytes) and FITC-dextran (vessel lumen). The frame size was typically 400 × 400 pixels (370 msper frame). x–y time series were taken to image pericytes and blood vesselsduring stimulation or micropipette puffing. Our earlier studies (3) used aframe rate of 5.9 Hz, but due to the properties of the instrument we used aframe rate of 2–3 Hz in the present study. The lower frame rate providedexcellent spatial resolution of the CVRs in an ensemble of small blood vesselsat the same time. However, this time resolution did not allow us to obtainrobust information about the latency time to 10% of the response. There-fore, in this study we used the time latency to 50% of the response.

Image Analysis. The analytical software was custom-made using MATLAB. Anaveraged image over time fromgreen channel was plotted. A rectangular ROIwith width of 4 μm was drawn perpendicularly across the vessel longitude(Fig. 1A). To minimize the interference with the black shadows of red bloodcells and minor vibration of the cortex, the rectangular ROI was averaged byprojection into one line for each frame, representing the profile of the vesselsegment at this frame. The profile line was plotted as a 2D image with the xaxis as number of frames (Fig. 1C, Upper). An active contour algorithm(Chan–Vese segmentation) was used to find the edges of the vessel, whichare indicated by the red curves (46, 47), and the time course of the measureddiameter was calculated according to the distance between the upper andlower red curves (Fig. 1C, Lower). Responding capillaries were defined asthose with a change of more than 2% of the initial vessel diameter. Thevessel response amplitude was defined as the highest peak amplitude after

C

BA

Fig. 7. Possible mechanisms of ATP-puffing-induced dilation and constriction. (A) The pial artery and p.a.s consist of endothelium surrounded by smoothmuscle cells (light red). As capillaries branch off the p.a., smooth muscle is replaced by pericytes (light blue) with heterogeneous morphologies across first-,second-, and higher-order capillaries (i.e., going from the p.a. to the venous side). (B) Fast and long-range conduction along arterioles and arteries viaelectrical conduction and the local Ca2+ wave. (C) Observed slow and low-range conduction of vasomotor responses emanating mostly from first- and second-order capillaries seem to involve signaling by diffusion. Both paracrine signaling along astrocyte end-feet and intracellular diffusion along putative gapjunctions can be envisioned.

Cai et al. PNAS | vol. 115 | no. 25 | E5803

NEU

ROSC

IENCE

PNASPL

US

Dow

nloa

ded

by g

uest

on

Oct

ober

9, 2

020

stimulation/puffing. The response latency was defined as the latency of half-maximal amplitude.

Drug Application. Upon completion of all surgical procedures, FITC-dextran(FD2000S; Sigma-Aldrich) was injected i.v. through the femoral vein cathe-ter to label blood serum and visualize the vasculature under the two-photonmicroscope (green color). In the micropipette ATP puffing study, the puffingsubstance was a mixture of 10 μM Alexa Fluor 594 (A-10438; Life Technol-ogies; red color) and 1 mM ATP (A9187; Sigma-Aldrich) dissolved in aCSF. Thesame method was used for the ATPγS (A1388; Sigma-Aldrich), UTP (U6875;Sigma-Aldrich), and αβ-methylene-ATP (M6517; Sigma-Aldrich) studies. Inthe preconditioned ATP puffing study using P2 receptor antagonist PPADS(P178; Sigma-Aldrich), aCSF containing 0.5 mM PPADS was used to superfusethe exposed cortex immediately after the dura mater was removed and toprepare both agarose and bathing fluid for the cranial window during im-aging. ATP puffing experiments occurred after at least 2 h of PPADS expo-

sure. The same procedure for PPADS application and exposure was used inthe subset of experiments in which cardiac arrest and cerebral ischemia wasinduced by i.v. application of 0.05 mL (200 mg/mL) pentobarbital.

Statistical Analysis. Responses are presented as mean ± SEM. P values arefrom one-way ANOVA with Tukey–Kramer post hoc test or unpaired Stu-dent’s t tests, as appropriate. P ≤ 0.05 was considered significant. All sta-tistical analyses were performed using MATLAB.

ACKNOWLEDGMENTS. We thank Krzysztof Kucharz for his inspiring adviceand professional help with polishing figures and videos and Alexey Brazhefor his kind help with improving the algorithm for video analysis. This studywas supported by the Lundbeck Foundation Research Initiative on BrainBarriers and Drug Delivery, the NOVO-Nordisk Foundation, the Danish Med-ical Research Council, and a Nordea Foundation Grant to the Center forHealthy Aging.

1. Abbott NJ, Patabendige AA, Dolman DE, Yusof SR, Begley DJ (2010) Structure andfunction of the blood-brain barrier. Neurobiol Dis 37:13–25.

2. Sakad�zi�c S, et al. (2014) Large arteriolar component of oxygen delivery implies a safemargin of oxygen supply to cerebral tissue. Nat Commun 5:5734.

3. Hall CN, et al. (2014) Capillary pericytes regulate cerebral blood flow in health anddisease. Nature 508:55–60.

4. Gould IG, Tsai P, Kleinfeld D, Linninger A (2017) The capillary bed offers the largesthemodynamic resistance to the cortical blood supply. J Cereb Blood Flow Metab 37:52–68.

5. Peppiatt CM, Howarth C, Mobbs P, Attwell D (2006) Bidirectional control of CNScapillary diameter by pericytes. Nature 443:700–704.

6. Kisler K, et al. (2017) Pericyte degeneration leads to neurovascular uncoupling andlimits oxygen supply to brain. Nat Neurosci 20:406–416.

7. Yemisci M, et al. (2009) Pericyte contraction induced by oxidative-nitrative stressimpairs capillary reflow despite successful opening of an occluded cerebral artery. NatMed 15:1031–1037.

8. Biesecker KR, et al. (2016) Glial cell calcium signaling mediates capillary regulation ofblood flow in the retina. J Neurosci 36:9435–9445.

9. Mishra A, et al. (2016) Astrocytes mediate neurovascular signaling to capillary peri-cytes but not to arterioles. Nat Neurosci 19:1619–1627.

10. Hill RA, et al. (2015) Regional blood flow in the normal and ischemic brain is con-trolled by arteriolar smooth muscle cell contractility and not by capillary pericytes.Neuron 87:95–110.

11. Fernández-Klett F, Offenhauser N, Dirnagl U, Priller J, Lindauer U (2010) Pericytes incapillaries are contractile in vivo, but arterioles mediate functional hyperemia in themouse brain. Proc Natl Acad Sci USA 107:22290–22295.

12. Wei HS, et al. (2016) Erythrocytes are oxygen-sensing regulators of the cerebral mi-crocirculation. Neuron 91:851–862.

13. Attwell D, Mishra A, Hall CN, O’Farrell FM, Dalkara T (2016) What is a pericyte?J Cereb Blood Flow Metab 36:451–455.

14. Hartmann DA, et al. (2015) Pericyte structure and distribution in the cerebral cortexrevealed by high-resolution imaging of transgenic mice. Neurophotonics 2:041402.

15. Mathiisen TM, Lehre KP, Danbolt NC, Ottersen OP (2010) The perivascular astroglialsheath provides a complete covering of the brain microvessels: An electron micro-scopic 3D reconstruction. Glia 58:1094–1103.

16. Wang X, Takano T, NedergaardM (2009) Astrocytic calcium signaling: Mechanism andimplications for functional brain imaging. Methods Mol Biol 489:93–109.

17. Fields RD, Burnstock G (2006) Purinergic signalling in neuron-glia interactions. NatRev Neurosci 7:423–436.

18. Kawamura H, et al. (2003) ATP: A vasoactive signal in the pericyte-containing mi-crovasculature of the rat retina. J Physiol 551:787–799.

19. Burnstock G (2016) An introduction to the roles of purinergic signalling in neuro-degeneration, neuroprotection and neuroregeneration. Neuropharmacology 104:4–17.

20. Frenguelli BG, Wigmore G, Llaudet E, Dale N (2007) Temporal and mechanistic dis-sociation of ATP and adenosine release during ischaemia in the mammalian hippo-campus. J Neurochem 101:1400–1413.

21. Schock SC, et al. (2007) Cortical spreading depression releases ATP into the extra-cellular space and purinergic receptor activation contributes to the induction of is-chemic tolerance. Brain Res 1168:129–138.

22. Hossmann KA, Traystman RJ (2009) Cerebral blood flow and the ischemic penumbra.Handb Clin Neurol 92:67–92.

23. Ames A, 3rd, Wright RL, Kowada M, Thurston JM, Majno G (1968) Cerebral ischemia.II. The no-reflow phenomenon. Am J Pathol 52:437–453.

24. Jensen LJ, Holstein-Rathlou NH (2013) The vascular conducted response in cerebralblood flow regulation. J Cereb Blood Flow Metab 33:649–656.

25. Dietrich HH, et al. (2009) Mechanism of ATP-induced local and conducted vasomotorresponses in isolated rat cerebral penetrating arterioles. J Vasc Res 46:253–264.

26. Kajita Y, Dietrich HH, Dacey RG, Jr (1996) Effects of oxyhemoglobin on local andpropagated vasodilatory responses induced by adenosine, adenosine diphosphate,and adenosine triphosphate in rat cerebral arterioles. J Neurosurg 85:908–916.

27. Ngai AC, Nguyen TS, Meno JR, Britz GW (2007) Postischemic augmentation of con-ducted dilation in cerebral arterioles. Stroke 38:124–130.

28. Ko KR, Ngai AC, Winn HR (1990) Role of adenosine in regulation of regional cerebralblood flow in sensory cortex. Am J Physiol 259:H1703–H1708.

29. Lämmer AB, et al. (2011) The P2 receptor antagonist PPADS supports recovery fromexperimental stroke in vivo. PLoS One 6:e19983.

30. Gaudin A, et al. (2014) Squalenoyl adenosine nanoparticles provide neuroprotectionafter stroke and spinal cord injury. Nat Nanotechnol 9:1054–1062.

31. Krueger M, Bechmann I (2010) CNS pericytes: Concepts, misconceptions, and a wayout. Glia 58:1–10.

32. Shih AY, et al. (2012) Two-photon microscopy as a tool to study blood flow andneurovascular coupling in the rodent brain. J Cereb Blood Flow Metab 32:1277–1309,and erratum (2013) 33:319.

33. Hill J, Rom S, Ramirez SH, Persidsky Y (2014) Emerging roles of pericytes in the reg-ulation of the neurovascular unit in health and disease. J Neuroimmune Pharmacol 9:591–605.

34. Grant RI, et al. (2017) Organizational hierarchy and structural diversity of microvas-cular pericytes in adult mouse cortex. J Cereb Blood Flow Metab 271678X17732229.

35. Masino SA, et al. (2002) Modulation of hippocampal glutamatergic transmission byATP is dependent on adenosine a(1) receptors. J Pharmacol Exp Ther 303:356–363.

36. Thyssen A, et al. (2010) Ectopic vesicular neurotransmitter release along sensory axonsmediates neurovascular coupling via glial calcium signaling. Proc Natl Acad Sci USA107:15258–15263.

37. Longden TA, et al. (2017) Capillary K+-sensing initiates retrograde hyperpolarizationto increase local cerebral blood flow. Nat Neurosci 20:717–726.

38. Chen BR, Kozberg MG, Bouchard MB, Shaik MA, Hillman EM (2014) A critical role forthe vascular endothelium in functional neurovascular coupling in the brain. J AmHeart Assoc 3:e000787.

39. Domeier TL, Segal SS (2007) Electromechanical and pharmacomechanical signallingpathways for conducted vasodilatation along endothelium of hamster feed arteries.J Physiol 579:175–186.

40. Filosa JA, et al. (2006) Local potassium signaling couples neuronal activity to vasodi-lation in the brain. Nat Neurosci 9:1397–1403.

41. Garland CJ, Dora KA (2017) EDH: Endothelium-dependent hyperpolarization andmicrovascular signalling. Acta Physiol (Oxf) 219:152–161.

42. Liu S, Agalliu D, Yu C, Fisher M (2012) The role of pericytes in blood-brain barrierfunction and stroke. Curr Pharm Des 18:3653–3662.

43. Flores-Soto E, et al. (2012) PPADS, a P2X receptor antagonist, as a novel inhibitor ofthe reverse mode of the Na+/Ca2+ exchanger in Guinea pig airway smooth muscle. EurJ Pharmacol 674:439–444.

44. Chen BC, Lee CM, Lin WW (1996) Inhibition of ecto-ATPase by PPADS, suramin andreactive blue in endothelial cells, C6 glioma cells and RAW 264.7 macrophages. Br JPharmacol 119:1628–1634.

45. Norup Nielsen A, Lauritzen M (2001) Coupling and uncoupling of activity-dependentincreases of neuronal activity and blood flow in rat somatosensory cortex. J Physiol533:773–785.

46. Chan TF, Sandberg BY, Vese LA (2000) Active contours without edges for vector-valued images. J Vis Commun Image Represent 11:130–141.

47. Getreuer P (2012) Chan–Vese segmentation. Image Process On Line 2:214–224.

E5804 | www.pnas.org/cgi/doi/10.1073/pnas.1707702115 Cai et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

9, 2

020